project-sharing_skill

Overview

This skill prepares organized, shareable packages of project files at three levels: summary PDFs, reproducible analysis bundles, and full traceability archives. It creates cleaned copies of notebooks, scripts, and selected data, produces README/MANIFEST documentation, and leaves your working directory unchanged. After package creation, all further work continues in the main project directory, not in the snapshot folder.

How this skill works

I ask which sharing level you need and a few context questions (audience, size limits, sensitive data). I build a dated sharing directory with an appropriate folder structure, copy and clean notebooks and scripts, include dependencies and documentation, handle or exclude sensitive data, and then compress the package if requested. I always return you to the original project directory and treat the shared folder as a read-only snapshot.

When to use it

• Share results with collaborators or reviewers
• Prepare supplementary materials for manuscripts
• Create reproducible packages for teaching or peer review
• Archive a completed project for compliance or handoff
• Submit data and code to repositories (Zenodo, Dryad)

Best practices

• Work on copies: never modify your main project when preparing a shareable package
• Choose the sharing level that matches the audience and purpose
• Clean notebooks: clear outputs, remove debug cells, ensure relative paths
• Document thoroughly: include README, MANIFEST, environment files, and data provenance
• Handle sensitive data deliberately: anonymize, exclude, or aggregate and document restrictions

Example use cases

• Level 1 (Summary): Export final notebooks to PDF with key figures and a brief README for non-technical stakeholders
• Level 2 (Reproducible): Provide cleaned notebooks, scripts, processed data, and an environment.yml for collaborators or reviewers
• Level 3 (Full Traceability): Package raw data, all processing steps, intermediate files, and full documentation for archival or regulatory needs
• Compress small packages as zip; use tar.gz and split for very large archives and document extraction steps
• Anonymize subject IDs via hashing before including datasets with personal identifiers